MGIS Projects - PowerPoint PPT Presentation

About This Presentation

MGIS Projects


This study examines data from the Revelstoke area ... Elizabeth Podgurny ... automata analysis, and calculates Shannon's Entropy (an urban sprawl index) ... – PowerPoint PPT presentation

Number of Views:1644
Avg rating:3.0/5.0
Slides: 56
Provided by: brett2


Transcript and Presenter's Notes

Title: MGIS Projects

MGIS Projects
Clickable Collage
Ulrich Aeschlimann
Environmental controls on glacier distribution
and the influence of spatial resolution of base
data on terrain analyses
  • In the context of global climate modifications,
    the monitoring of glaciers is a key instrument to
    quantify changes and to predict any resulting
    adverse impacts on alpine environments. This
    study examines data from the Revelstoke area
    (Glacier National Park, British Columbia) and the
    Kananaskis region (Peter Lougheed Provincial
    Park, Alberta) to address three issues that are
    paramount in any GIS-based, elevation model
    operations. The three modules of the study
    include Glacier parameter inventory, DEM -
    Resolution considerations, Climate Modeling.
    The glacier parameterization module extracts and
    compares elevation related factors (e.g., Mean
    elevation, ELA) as well as topographic parameters
    such as slope angles, aspect and curvature. The
    comparison of the glacier parameters of the two
    study areas indicates significant differences in
    the average glacier mean elevation and the
    average ELA. However, there is no statistical
    difference in average glacier area, length, shape
    index, and overall slope gradient between the two
    study areas. The second module, DEM resolution
    considerations, compares elevations extracted
    from different grid resolutions (30 m, 100 m, 200
    m, and 500 m) to determine if certain parameters
    (e.g., mean elevation) are less problematic if
    obtained from low resolutions than other
    parameters, for example minimum and maximum
    elevations. The assessment shows that the average
    glacier mean elevation is similar (a maximum of 7
    m difference) between values extracted from a 30
    m grid and values extracted from a 500 m grid.
    However, the maximum and minimum elevations are
    considerably different between the 30 m and 500 m
    grids, particularly in the Kananaskis study area.
    As such, caution is recommended when using
    glacier parameters obtained from low resolution
    DEMs. In particular, grid resolutions of 500 m
    and larger may cause considerable errors in
    parameters measured. The third module, climate
    models for the two study areas, explores general
    ways of evaluating the role of climate under
    different climatic regimes. The chosen method
    includes extrapolation of 30-year data sets from
    climate stations using linear regression
    analysis. The climatic parameters assessed in
    this study are temperature and precipitation
    (i.e., snow depth). For the Revelstoke area, the
    temperature/elevation results indicate that there
    was a decrease of 0.53 C in the mean July
    temperature with at every 100 m elevation
    increase (intercept 20.3 C). The mean January
    temperature decreased by 0.48 C at every 100 m
    elevation increase (intercept 8.4 C). Due to a
    lack of alpine climate data in the Kananaskis
    area, the model was based on a constant gradient
    of 0.6 C/100 m with intercepts of 22.3 C for
    July and 11.4 C for January. The intercepts were
    extrapolated from climate stations close to the
    study area. The Revelstoke data indicates an
    overall increase of 72 cm in snow depth per 100 m
    increase in elevation (intercept 89 cm). In the
    Kananaskis area the average increase in snow
    depth is 29 cm per 100 m increase in elevation
    (intercept 132 cm).

Adriana Aguilar
A Multi Criteria Evaluation approach using Fuzzy
Set Theory
  • A computerized decision support tool implementing
    a multiple criteria evaluation approach using
    fuzzy set theory is programmed in Arc/Info as a
    prototype. To aid and evaluate lynx habitat
    selection in Southern Alberta by developing a
    surface of possibility, following three different
    approaches and comparing the results.
  • The prototype presented herein has been
    programmed in Arc/Info macro language (AML)
    following a modular approach and presented in a
    series of easy to use menus.
  • By combining geographic information systems (GIS)
    and multiple criteria evaluation (MCE)
    capabilities and in the context of analyzing
    geographic data the union of these methodologies
    appears to be instrumental in the design of
    efficient tools for spatial decision-making.

Catherine Bow
A GIS Based Study of Address and Postal Code
Location and Street Network Files and their use
in Medical Geography Research
  • Medical Cardiac Catheterization Data supplied
    from the APPROACH (Alberta Provincial Project for
    Outcome Assessment in Coronary Heart Disease)
    database was reviewed and analyzed to determine
    the amount of error propagation inherent within
    the database. The error was found by comparing
    the address and postal code information in the
    database to the postal code and address
    information found in the 2001 Edition City of
    Calgary telephone book. A 16.32 percent error
    rate was found, excluding missed values and minor
    spelling errors. The database was corrected,
    postal code converted to find the longitude and
    latitude of the postal code, and geocoded into a
    GIS (Geographic Information Systems), namely
    ESRIs ArcView 3.2. Two street network files were
    compared, the City of Calgary 2001 file and the
    Statistics Canada 1996 file. The address and
    postal code locations were found for both of
    these files in meters and subtracted. In SPSS
    (Statistical Package for Social Scientists),
    scattergrams and histograms were created to
    compare these two files. It was found that the
    Statistics Canada street network file was less
    accurate and had more errors within the database.
    The City of Calgary street network file was
    deemed to be very accurate as it was used and
    updated daily within the City. The subtraction of
    the address and postal code led to similar
    results in terms of the mean and standard
    deviation, but the Statistics Canada file created
    an unusually linear pattern at the 25 and 25
    meter x and y-axis. This linear pattern is due to
    the block-face representative postal code point
    location of 20 meters from the centre of the
    street. Depending on the application, the postal
    code was deemed to be a fairly good
    representation of the address location.

Wesley Burrows
A 3-D Steady-state groundwater model for West
Nose Creek watershed
  • A 3-dimensional, steady-state, numerical
    groundwater model for the West Nose Creek
    watershed is proposed in this project. Hydraulic
    conductivity values are derived from summaries of
    pump test data from domestic and stock water
    wells. Spatially distributed hydraulic
    conductivity is interpolated from water well
    point estimates and imported to MODFLOW. For
    this, a loose-couple between ArcGIS and MODFLOW
    is implemented through Visual Basic and
    ArcObjects and is demonstrated to perform
    correctly. Calibration of the numerical model was
    not possible, as convergence was not achieved. An
    inability to model stream leakage and complex
    heterogeneities in hydraulic conductivity fields,
    has led to an ill-conditioned problem and
    ultimately non-convergence of this
    finite-difference solution.

Bruce Burwell
Implementing a Geographic Information System
Using ESRIs Geodatabase Model A Case Study in
Archaeological Research
  • This research demonstrates the conceptual design
    and system implementation of a Geographic
    Information System for archaeological research
    using ESRIs (Environmental Systems Research
    Institute) Geodatabase model. The Geodatabase
    model uses an Object-Oriented approach to
    database design in comparison with other data
    basing techniques which utilize
    Entity-Relationship design methodology. Based on
    research conducted in current practices of data
    modeling and implementation in archaeology, a
    spatial data model was created using Microsofts
    Visio 2000 drawing and modeling package. The
    model was then imported into ArcGIS and populated
    with data available from archaeological research
    previously conducted in Yemen. To complement the
    database, a wide variety of public domain image
    and vector data was also processed and modeled in
    the database. An application was then created
    that would facilitate a preconceived workflow for
    database updating, data mining, mapping and
    reporting. The result is a fully functional GIS
    (geographic information system) database
    management system with supporting model and
    documentation. This project also provides an
    effective methodology for designing and deploying
    an archaeological data model and proves that
    ESRIs object oriented database is superior to
    previous methods of GIS development.

Nicola Bywater
Cumulative Effects of Hurricane Iris (2001) on
Belizean Black Howler Monkey Habitat
  • This study integrated remote sensing and
    geographic information systems (GIS) to quantify
    landscape fragmentation and habitat change
    resulting from Hurricane Iris (2001) in the Stann
    Creek District of Belize. There are few studies
    that integrate remote sensing, GIS, primates, and
    tropical forests. Ground surveying was performed
    in January 2003, which identified 15 major land
    cover types in this area. The ground survey data
    was used to classify a pre hurricane Landsat 5
    image and a post hurricane Landsat 7 image.
    Habitat change was quantified by comparing the
    pre and post hurricane classified images using
    FRAGSTATS version 3.3 software. Class and
    landscape level metrics were used to quantify
    habitat change and fragmentation. It was found
    that the landscape has become more fragmented,
    specifically increase in patch numbers, increase
    in total edge, decrease in average patch size,
    decrease in patch extent, significant loss of
    core area, decrease of functional connectivity,
    increased interspersion, and an increase of
    aggregation for some patch types. These findings
    could have significant implications for howlers,
    particularly a loss of habitat quality and

Christopher Caschera
Multiresolution Wavelet Based Satellite Image
Fusion Application to examining Arctic sea ice
  • In this document the application of wavelet
    theory to fuse disparate multiresolution
    satellite images is performed for evaluating the
    results against Canadian Ice Service ice chart
    derived estimates. The primary images will be
    RADARSAT-1 ScanSAR products (100m) from the
    Canadian Data Processing Facility (CDPF), NOAA
    AVHRR Thermal band 4 (1100m) provided by the
    National Snow and Ice Data Center (NSIDC), and
    passive microwave from the DMSP SSM/I at 25km
    resolution. The temporal scale for this study
    ranges from Julian Day 137 (May 17) to 204 (July
    23), 2002. The study area is located in Lancaster
    Sound between Borden Peninsula, Baffin Island and
    Cape Warrender, Devon Island (748'N, 8113').
    The images are decomposed into their multi-scale
    edge representation, using Daubechies wavelet
    transformation. The fusion process is built by a
    pixel by pixel selection of the coefficients with
    maximum amplitude. Finally, the fused image is
    computed by the appropriate reconstruction
  • The wavelet image fusion results in a
    non-redundant image representation providing a
    photointerpretive and statistical assessment of
    wavelet based fusion of sea ice in the study
    area. Results indicate that the fusion of
    RADARSAT-1 ScanSAR to NOAA AVHRR is comparable
    and in some instances provides additional data
    over Canadian Ice Service (CIS) ice charts.
    Specifically, fusion conducted on JD 180 and 200
    illustrate the strength and applicability of this
    method. The fusion of RADARSAT-1 ScanSAR to an
    SSM\I algorithm of sea ice concentration
    (Bootstrap) illustrated no benefit over current
    methods employed by the CIS. Pixel size appears
    to be a major constraint for the successful
    fusion of images based on this method.

Sam Coiro
Spatial Analysis of Enterprise Customer Data
  • The construction of trade areas for retail
    selling channels bound by a fixed location in
    space, such as the traditional bricks-and-mortar
    selling unit, has been extensively reviewed in
    literature over the past 50 years. Although the
    catalog selling channel is the topic of much
    literature, attempts to describe, explain, and
    predict trade areas of such channels have yet to
    be made. Unlike traditional channels, catalog
    ones are not bound by fixed spatial locations and
    thus the methodologies used to derive
    bricks-and-mortar trade areas cannot be directly
    applied to catalog channels. This paper proposes
    the use of trend surface analysis (and the
    creation of visualization maps) to identify
    catalog trade areas at varying scales of
    geography. Four product division groups are
    studied and results indicate that for the most
    part, the trade area formations are dependent
    directly on the type of product in questions. In
    general, trend surfaces created for hard-line
    merchandise (e.g. tools, furniture, etc.) heavily
    underpredict transaction amounts across space.
    Trend surfaces created for soft-line merchandise
    (i.e. clothing, bedding, etc.) heavily
    overpredict transaction amounts across space.
    Regardless of the observed inadequacies however,
    the products of this paper are excellent
    first-steps that retail marketers can use to
    render the most return on investment possible
    from catalog campaigns, distributions, and

Jon Connick
Benchmarking a Geospatial Store ArcSDE and
Oracle Spatial
  • This manuscript describes an evaluation of the
    spatial storage capabilities of ESRIs ArcSDE and
    Oracles Oracle8i Spatial. The prime objective of
    this evaluation is to understand how to implement
    and manage a spatial store in an Oracle DBMS
    using ArcGIS as the client. The evaluation was
    completed through the implementation of the
    SEQUOIA 2000 Storage Benchmark and Paradise
    Geo-Spatial DBMS Benchmark.
  • The benchmarks provided insight into performance
    differences between ArcSDE and Oracle8i Spatial
    storage and indexing options, in addition to
    overall functionality of the ArcGIS client in
    completing the queries. A chapter devoted to
    ArcSDE and Oracle8i Spatial functionality is
    provided to outline the common and differing
    aspects of both softwares.
  • This study showed that data stored as ArcSDE
    Binary Geometry - Long Raw datatype, performed
    better overall in completing the queries and
    required less storage table space than the ArcSDE
    Binary Geometry - BLOB datatype and all other
    Oracle8i Spatial options.

Kathleen Donovan
Snowmelt Run-off modeling of Prairie Depressions
  • Understanding and quantifying snowmelt retention
    in depressions is an important component of
    modeling their role in groundwater interactions
    such as infiltration, soil moisture recharge, and
    the overall hydrologic properties of the
    watershed such as regional aquifer recharge, and
    stream discharge. The potential volume of
    snowmelt runoff was estimated using average snow
    water equivalent, depression area estimates, and
    depression depth measurements. The area of
    surface water runoff retained in the depressions
    after snowmelt was estimated from supervised
    classification of infrared aerial photographs
    with an accuracy of 15 compared to the detailed
    elevation surfaces, and 10 compared hand
    digitized polygons of the depressions on the
    photographs. Using 10 detailed elevation
    surfaces, the relationship between depression
    volume (V) and depression area and depth (Ah) was
    defined as V0.515(Ah). The error for this
    relationship compared to the volume estimated
    from the elevation surfaces was 10. The
    classified area and measured depth of 101
    depressions were applied to this equation to
    determine the overall relationship between
    depression area and volume. The relationship
    between depression area and volume was defined
    using least squared regression as V0.055
    (A)1.17. The cumulative error associated with
    this model is estimated to be approximately 25.
    Application of the volume model to all the
    classified depressions indicated that 40 to 70
    of the snow water available for runoff was
    retained by depressions. This assessment
    indicates that depressions in the West Nose Creek
    basin provide a significant contribution to the
    storage component of the water balance in the

Rebecca Evans
GIS and on-line learning pedagogical issues and
tutorial development
  • This project is a study of Geographic Information
    Systems (GIS) and online learning. The study is
    presented in two parts. Part One, Theoretical
    Issues (GIS and Online Learning Pedagogical
    Issues), is an examination of the broader,
    conceptual issues surrounding online GIS
    learning. It is argued that online,
    computer-based tutorials are an effective way to
    teach the concepts, applications, and technical
    skills required of a competent GIS user. However,
    it is important that GIS online learning
    materialsespecially those designed for use in
    higher educationbe based on a strong foundation
    of psychological learning theory. Thus,
    constructivism and the various approaches to
    constructivist learning and teaching are
    discussed. It is demonstrated how the use of
    technology, in combination with constructivist
    learning principles and the careful consideration
    of instructional design can produce quality,
    pedagogically sound, online GIS learning
    materials. Part Two, Implementation (An Online
    GIS Tutorial Development Project), is the
    practical application of the concepts, theories,
    and approaches regarding GIS and online learning
    discussed in Part One of this report. A series of
    GIS tutorials that demonstrate the use of GIS as
    a cartographic tool for thematic mapping are
    developed and implemented in an online
    environment. The tutorials are customized for use
    in an existing course offered in the Department
    of Geography, University of Calgary, and
    incorporate the key elements of quality online
    pedagogical principles and instructional design.
    Two primary software products, ArcView GIS and
    WebCT, are utilized to accomplish this goal.
    Preliminary results are obtained by having two
    anonymous volunteers test the tutorials and offer
    their comments and recommendations. The results
    of these tests indicate a high degree of success
    and it is concluded that the GIS tutorials hold
    considerable promise as an instructional/learning

Neil Farries
Study Wildlife Overpasses Crossing TransCanada
  • This project begins with an overview of the
    history and operation of Cellular Automata (CA)
    as a tool for predictive change modeling. Several
    diverse projects from a number of disciplines are
    presented as examples of the vast amount of
    research that has been completed in this area.
    Next, an application project is presented in
    which the CA modeling capabilities of the Idrisi
    CELLATOM module are tested. This project entails
    a predictive model of Mountain Pine Beetle (MPB)
    infestation in the Bow Valley Corridor near
    Canmore, Alberta.

Wenonah Fraser
Airborne laser scanning for measuring trees in
the boreal forest theoretical issues and
  • Airborne laser scanning is a developing
    technology with many application possibilities.
    The first part of this document focuses on
    theoretical issues about using airborne laser
    scanning (also known as lidar) for assessing and
    monitoring forests. A review of forest structure
    is followed by an introduction to light detecting
    and ranging techniques and characteristics, with
    a review of applications in other fields. After
    explaining some of the challenges unique to
    lidar, issues regarding grid cell sampling are
    explored. Finally, potential interpretation
    techniques to identify individual trees by
    filtering are discussed. The second part
    demonstrates an application using lidar data.
    First and last returns from seventeen plots in
    the boreal forest in northwestern Alberta were
    used to generate estimates of tree heights.
    Kriging was used to interpolate surfaces of the
    tree canopy. The estimated tree heights were
    compared to tree heights determined from
    high-resolution aerial photographs. Although the
    relative tree heights within each plot compared
    well, lidar underestimated tree heights somewhat.
    Others have found similar results with different

Roger Gauvin
Object-based Terrain Classification of a High
Resolution Elevation Model
  • Terrain is accredited as one of the largest
    factors controlling surface and atmospheric
    processes on the earth (Hutchinson and Gallant,
    2000). Terrain landforms are defined to be the
    resulting morphology and characterization of
    slopes and slope patterns caused by physical
    processes on the earth (Whittow, 2000). Terrain
    itself can be said to be the spatial arrangement
    of landforms into a repeatable pattern. These
    patterns of terrain can be explained by the
    catena concept of landscape soil process was
    first described by Milne (1935) and later
    expanded with Spreight (1974) as the realization
    that sequences of slope forms were intrinsically
    related with soil series in a predictable way.
    Before the advent of spatially enabling
    technologies such as Geographic Information
    Systems (GIS), the science of geomorphometry
    involved either qualitative description or
    quantitative field measurements. Purely
    qualitative descriptions are problematic due to
    the lack of standardization of descriptive labels
    required to extend global classification.
    Quantification of landscape attributes required
    for terrain classification prior to the practice
    of computer automation required detailed field
    measurements involving laborious human
    intervention and limited analytical capability.
    Soon after the common adoption of the computer
    representations of terrain in the digital
    elevation models in the late 1980s, earth
    scientists and ecologists became intrigued with
    the potential of using elevation models to create
    specific topographic attributes that have a
    predictable relationships to both biotic and
    abiotic landscape processes. Better understanding
    of these key processes was to generate interests
    in applications for hydrology, soil conservation
    and forestry creating benefits for improved
    environmental management. In the 1990s the
    advent of high resolution Digital Elevation
    Models (DEMs), created with Light Detecting and
    Ranging (LIDAR), Inferometric RADAR or automated
    softcopy photogrammetric techniques, created new
    challenges and opportunities for the extraction
    of detailed landscape information for
    applications using terrain modeling techniques.
    In this review, quantitative methods that employ
    the use of GIS and spatial analysis techniques
    for the purpose of terrain analysis will be
    highlighted. Geostatistical exploration with
    spatial statistics of the digital terrain models
    attributes is argued to be essential for
    determining the structure, domain and range of
    these attributes. The determination of the
    spatial domain and range of terrain derivatives
    are necessary to approximate the scale of
    landform entity phenomena and how they might
    relate to complex landscape processes. As it is
    widely accepted that different landscape
    processes operate at different scales and the use
    of hierarchical object-based schemes may be the
    best method to model these complex relationships.
    This review will outline some of the theoretical
    issues of scale, hierarchies and the practical
    realities of designing a spatially aware modeling
    framework for modeling. This research area began
    with simple morphometric characterization of
    landscape and terrain, but is now moving to
    incorporate contextual methods that may better
    describe the spatial arrangement inherit of
    hierarchical schemas where a variety of landscape
    scales and processes interact. Recently, there
    has been much interest in using automatic
    landform segmentation and classification for
    practical applications in geotechnical
    engineering, agricultural and the environmental
    field. With these advancements, large amounts of
    data (and large tracts of landscape) have the
    potential to be analyzed. The question is as
    these new technologies become adopted will the
    statistical quantification in this field
    accelerate faster than the semantic labels to
    make practical sense of the relationships between
    processes and forms?

James Gebert
ArcGIS migration to support upstream oil and gas
  • This project was undertaken to examine the
    migration from ArcView 3.x to ArcGIS 8.x in
    petroleum exploration and production (EP)
    companies. The major focus was to investigate how
    previously developed customization in ArcView 3.x
    can be implemented in ArcGIS 8.x. A custom tool
    was developed to automatically plot and symbolize
    wells from a standard petroleum industry
    database. The development of this tool
    demonstrates one aspect of ArcGIS customization.
    Other customization options for the creation of
    an oil and gas GIS are also summarized. As would
    be expected, petroleum specific functionality
    developed for ArcView 3.x can be replicated in
    ArcGIS 8.x, although at the considerable expense
    of time and effort. Development is more difficult
    in ArcGIS 8.x because of the increased complexity
    of the system architecture. However, this
    increased complexity allows ArcGIS 8.x to be
    customized to a much greater extent than ArcView
    3.x. A major benefit of migration is the
    geodatabase, which is a more robust format for
    the storage and maintenance of spatial data. The
    collaborative development of petroleum specific
    data models may facilitate the loading of data
    into the geodatabase, and make data access more
    efficient. There are limited short-term benefits
    to migration for the exploration business unit
    although the long-term benefits may be
    significant. The ability to customize ArcGIS 8.x
    will no doubt lead to the future development of
    additional tools that will directly benefit the
    exploration business unit. This will allow GIS to
    make incremental progress in the transition from
    its current role as a tool of data visualization
    towards a future role as a tool of spatial
    analysis and modeling.

Nicole Hopkins
Analysis of wetlands in Scotty Creek basin, NWT,
using IKONOS and Landsat imagery
  • Wetlands are an important part of the hydrology
    of the Scotty Creek basin, near Fort Simpson,
    North West Territories, because of their ability
    to hold water which affects the spring flood on
    the Liard and Mackenzie Rivers. The 30-meter
    resolution Landsat Thematic Mapper and the finer
    resolution IKONOS imagery (4 meters) were used to
    classify the Scotty Creek basin into six classes
    lakes, channel fens, wetlands, coniferous,
    deciduous and mixed forest. The percentage of the
    basin occupied by wetlands was also determined
    using commercially available computer programs.
    The IKONOS imagery was only available for the
    east portion of the basin. Therefore, the
    objective of the study was to classify the east
    portion of the basin using the finer resolution
    IKONOS imager and classify the entire basin with
    the Landsat image using the common training
    sites, compare the results and evaluate the
    accuracy. The overall accuracy for the hydrology
    landcover classes, which includes lakes, channel
    fens, and wetlands, was 78 for the IKONOS image
    and 67 for the Landsat classification. Both
    classifications resulted in confusion between the
    conifer and mixed vegetation classes. The
    accuracy for the Landsat classification was lower
    because of greater confusion between the conifer,
    channel fen and wetlands classes due to the
    coarser resolution of Landsat imagery. The IKONOS
    classification resulted in smaller patches (i.e.,
    interconnected pixels having the same class)
    which were closer together than the Landsat
    patches and determined that 41 of the east
    portion of the basin was occupied by the
    hydrology classes whereas the Landsat showed that
    46 of the basin was occupied by the hydrology
    classes. The IKONOS classification resulted in a
    greater accuracy and provided a better
    representation of the hydrology and connectivity
    of wetlands in the Scotty Creek Basin.

Luigi Iannuzzi
Socio-economic application of geographic
information systems, calculations and spatial
analysis of the Human Development Index for
  • In 1990 the United Nations Development Program
    published the first world Human Development
    Report. This report refocused the attention of
    developmental practitioners on people. It also
    reminded us that economic growth is only a means
    to improve people's lives and not an end in
    itself. Human development is and should be the
    goal for any development. The Human Development
    Index (HDI) was developed as a new aggregate
    indicator, which would make it possible to
    compare the level of development of individual
    countries. It is always difficult to construct
    such aggregate indicators for it requires a
    number of theoretical questions to be resolved
    and open criticism to be answered (Banuri, 1994).
    The HDI has been undergoing this process over the
    last five years. There are many different ways
    that such an indicator can be constructed and if
    it is to be used as a means of comparing
    different countries throughout the world, the
    availability and reliability of data must be
    taken into account. After the methodology of
    calculating the indicator was developed, its
    three basic components remained unchanged. One
    was the per capita gross domestic product (GDP)
    which is an indicator of the economic level of
    the country, this being one of the basic
    preconditions for human development. When
    calculating the HDI it is adjusted for parity of
    purchasing power. The second component for
    directly measuring human development is life
    expectancy at birth as an average for both sexes,
    and the third is the level of literacy of the
    adult population, recently supplemented by the
    average period of school attendance. This first
    Development Index for Honduras at the municipal
    level was produced for the Field office of the
    United Nations Development Programme in
    Tegucigalpa Honduras. This will be the base
    document to be used in the development of this

Keila Johnston
Archeological Predictive Model for the Souris
River Basin Region of Saskatchewan
  • This study employs GIS and statistical analysis
    in an effort to construct an archaeological
    predictive model for determining potential site
    locations in the Souris River Basin of
    Saskatchewan. Archaeological predictive modeling
    involves the use of environmental correlates with
    site presence or known human behaviour patterns
    in an attempt to locate likely areas of site
    presence. Inductive predictive modeling, or
    environmental correlates of site location are
    used to determine the probability of site
    occurrence throughout the Souris River region.
    Chosen environmental variables include distance
    to water, elevation, slope and land cover.
    Predictive modeling can prove to be a valuable
    aid in cultural resource management work because
    the time, money and manpower required to
    undertake extensive surveys can be greatly
    reduced. Statistics used in this study include
    nearest neighbour analysis, chi-square analysis
    and Dempster-Shafer theory. Nearest neighbour and
    chi-square analysis were used to determine the
    spatial patterning found in the region and to
    determine the association of archaeological sites
    with environmental variables. The Velief Module
    in Idrisi employs Dempster-Shafer theory and was
    used to construct a predictive model for the
    area. The Relative Operating Characteristic (ROC)
    Module in Idrisi was used to validate the results
    of the Belief Module. The sites of the Souris
    River Basin are clustered. A clustered pattern
    can be indicative of an abundance of good
    resources. Distance to water, elevation, slope
    and land cover are correlated with site location.
    Dempster-Shafer theory can be used to accurately
    reconstruct site location in the Souris River
    region. High probability areas are areas of
    possible site location, not areas of definite
    site location.

Christopher Jordan
Client Server Feature Streaming in ArcGIS 8
  • Large corporations such as those in the Petroleum
    industry amass a vast quantity of spatial
    information. Data accuracy and currency is
    paramount in these organizations, with
    problematic data frequently leading to poor
    decision-making and a substantial loss of
    capital. An efficient process of maintaining data
    currency over the web has been developed and is
    appropriately named Feature Streaming.
    Web-based Feature Streaming allows limited data
    to be delivered from a Web Server to a client
    computer only when required, thereby resulting in
    increased efficiency. Just as web-based Feature
    Streaming effectively removes the internet
    bottleneck, Selective Client-Server Feature
    Updating removes the data transfer bottleneck
    associated with the corporate intranet. Visual
    Basic code has been developed using ArcObjects
    and Microsofts Component Object Model to enable
    Selective Client-Server Feature Updating in
    ESRIs ArcGIS 8.3 software. This program has been
    developed for EnCana Corporation to be used with
    ArcSDE. The code facilitates transparent data
    currency by regularly updating client machines
    with modifications, additions and deletions to
    the tables of spatial datasets. By eliminating
    the redundancy involved in the updating of entire
    records, the bottleneck created by the intranet
    is effectively removed, allowing massive datasets
    to be refreshed on client machines without delay.

Hejun Kang
Remote Sensing and GIS Applications in Lynx
Habitat Suitability Modeling
  • Geographical Information Science (GIS), with its
    versatility and potential in addressing
    ecological issues, has had numerous methods for
    modeling wildlife habitat use and suitability.
    The first goal of this research was to identify
    the habitat preference of both lynx and snowshoe
    hare, and to determine relationships between the
    two species. Secondly, predict the lynx
    occurrence through different modeling methods
    Logistic regression, Multiple-Criteria
    Evaluation (MCE), and Dempster-Shafer, and then
    compare the prediction power of each model. Based
    on the overall accuracy, the best model is
    logistic regression (74.51), then the MCE
    (unequal weights), MCE (equal weights), and
    finally the D-S model (29.90). However, using
    only the users accuracy, the rank is completely
    reversed the best one is the D-S model (100),
    and the worst one is the logistic regression
    model (27.07). There are two main factors
    affecting the reliability of models. One is the
    combination of absence points, which inflated the
    overall accuracy of both the logistic regression
    model and the MCE (unequal weights), and
    decreased that of the other two models. Another
    is the choice of the threshold value. Logistic
    regression, which is more objective, can help
    decide how to choose prediction factors for both
    the MCE and D-S models. Future research should
    increase sampling on areas with higher slopes for
    both the MCE and D-S models. For the logistic
    regression model, spatial autocorrelation should
    be incorporated to increase the model performance.

Maria Komierowski
Using GIS to Identify Wind Erosion Risk and to
Recommend Sustainable Land Management Practices
in the Taber Irrigation District of Southern
  • Soil erosion is a predominant problem in Southern
    Alberta, especially with the strong chinook winds
    that are characteristic in the area. Additional
    problems exist on fields where specialty crops
    (potatoes, sugar beets, dry beans) are grown.
    Soil erosion poses a threat to human health and
    safety by air-borne particulate matter as well as
    reducing visibility on highways. Soil erosion
    also poses a threat to the sustainability of
    agricultural land and the agricultural industry.
    This work presents a state of the land report
    that identifies areas that are of concern to wind
    erosion, as well as locating areas that allow for
    sustainable expansion of the potato industry.
    This report also discusses various methods and
    practices that could reduce the risk of a wind
    erosion event. Using a Pairwise Comparison Matrix
    and ArcView 3.2, factors that affect wind erosion
    were identified. These factors were then used to
    locate the aforementioned areas of interest. The
    study found that areas of concern and highest
    priority for mitigation are those that have
    easily erodible soil textures (sands and clays)
    but have highly suitable land for potato
    production. It was determined that 48 of the
    Table Irrigation District fell within this
    category. Areas suitable for expansion included
    those that had inherent soil properties that have
    a reduced risk of wind erosion (loams and silts),
    had land suitable for potato production, and were
    in proximity to land with existing irrigation
    rights. Mitigation practices such as proper
    education, residue covers, shelterbelts, and
    emergency control methods were identified.

Dae Won Kwon
An Analysis of the Impacts on Travel Behavior
during the Transit Strike in Calgary using
Discrete Choice Methods
  • Mode choice models are used to analyze and
    predict the choices that individuals or groups of
    individuals make in the selecting the
    transportation modes that are used for particular
    types of trips. This study describes an
    investigation of the impacts on mode choice
    travel behavior during the transit strike in
    Calgary using logit models. A utility function is
    formed which includes attributes that describe
    socioeconomic characteristics of households and
    individuals. The binary logit models and
    multinomial logit models based on this utility
    function are developed in cases of Home Based
    Work trip and Home Based Other trip before and
    during the transit strike. The developed logit
    models are evaluated using an individual level
    (disaggregate) data set and a zone level
    (aggregate) data set. The parameter estimation
    and model evaluation are performed using GIS
    software (TransCAD). The models that result and
    the model evaluation provide indications of the
    impacts on travel behavior of Calgarians during
    the transit strike.

Dana Lampi
Aboriginal GIS
  • Aboriginal groups in Canada are taking an
    increasingly active stance in the definition and
    control of their traditional territories. This is
    occurring through two major processes land claim
    settlements and resources management /
    co-management. Each of these activities has a
    large spatial component and Aboriginal
    communities are turning to Geographic Information
    Systems (GIS) to address these issues. GIS is
    being used as a tool to provide a non-western
    view of the traditional territories in the face
    of development and is not only building capacity
    within Aboriginal communities but also providing
    community members with an in depth look at their
    culture, both past and present. The second part
    of this paper provides the course materials for
    an Aboriginal GIS course to be offered as part of
    the GIS Applied Degree program at the Southern
    Alberta Institute of Technology (SAIT). The
    purpose of the course is to introduce students to
    the different ways that GIS is being used by
    Aboriginal groups and to provide students with an
    introduction on GIS implementation in Aboriginal

Joe Lee
An exploration into the use of digital cameras
for remote sensing studies of live vegetation in
Grasslands National Park
  • The movement of animals across a landscape has
    been of interest to landscape ecologists for
    years. One way of studying this movement is
    through the development of simulation programs.
    These programs attempt to mimic the movement of
    animals in real life through the use of various
    algorithms and models. More recently, the
    development of object-oriented models for
    software design has lead to new approaches to
    this problem. Artificial intelligence has
    provided the framework for intelligent agents,
    programs that actively work to achieve a goal
    using knowledge and resources at its disposal.
    This project examines the complexities involved
    in the design of such simulations by developing
    an animal movement simulation that attempts to
    capture the effects of habitat patch
    configuration upon animal movement.

Nancy Lee
A Spatially Explicit, Object-oriented Simulation
Model of Animal Movementulation Model of Animal
  • The use of digital photography to gather data for
    remote sensing studies on vegetation has a number
    of advantages over data collected in air or
    space. Digital cameras are small and easily
    transportable which makes the technology readily
    available for fieldwork. Likewise it is
    relatively inexpensive as compared to airphotos
    and satellite imagery. Photos can be taken of any
    areas at any time and can instantly be
    transferred to a computer for analysis with
    remote sensing software packages. Unfortunately,
    the majority of digital cameras take photos in
    the RGB wavelengths only. This is limiting with
    respect to vegetation analyses because the
    spectral reflectance curves of live and dead
    vegetation and bare soil are note very different
    in the VIS wavelengths. Consequently, procedures
    such as PCA, supervised classification,
    unsupervised classification and ANNs performed
    poorly at classifying these features whose
    spectral information was taken with the VIS
    spectrum. Reflectance information for vegetation
    in the VIS wavelengths was not sufficient to
    enable good spectral separability of the features
    in data space. The average blue radiance values
    for live grass were extracted from the
    classifications and plotted against the total
    green biomass clipped and weighed for the plots.
    Additionally the percent area of live grass and
    forbs was plotted against the total green biomass
    weights. No relationships were found. Vegetation
    indices such as the NDVI and the NDVI with the
    blue and green bands (referred to as the NBVI)
    were done to extract the live vegetation
    features. The values for the NDVI calculations
    were collected using s spectroradiometer and the
    MODEL function in PCI Works. These values were
    also plotted against the total green biomass
    weights and no relationships were seen. However,
    sample sizes were too small to allow solid

Kaidong Li
Case-based reasoning, traffic safety and GIS
  • This paper reviewed previous studies on the
    application of geographic information systems
    (GIS) and case-based reasoning (CBR) in traffic
    safety research, and demonstrated the methodology
    and results of a study that used GIS and CBR to
    analyze and manage traffic safety at light rail
    transit (LRT) grade crossings in the City of
    Calgary. GIS have generated an increasing amount
    of attention in the traffic safety community due
    to their capabilities for storing and analyzing
    spatially distributed data, such as traffic
    collisions. The information-retrieval methods,
    adaptation capabilities and learning processes of
    the CBR systems are considered more advanced than
    other database and reasoning systems. Therefore,
    the integration of CBR and GIS is highly likely
    to be beneficial in developing an intelligent
    system for traffic safety analysis. In this
    project, traffic collision history and
    site-specific data at Calgarys LRT grade
    crossings were manipulated and analyzed in
    ArcGIS. This data was then put into a CBR system,
    eGain Knowledge, to establish a case base for
    future LRT crossing studies. An example from the
    case base was provided.

Travis Logan
Large mammal habitat modelling Logistic
regression and model selection using GIS
  • Cougar habitat suitability was modeled as a
    function of a number of landscape variables,
    describing features of topography, vegetation, as
    well as prey and competitor species densities.
    Geographic Information Systems (GIS) and Remote
    Sensing (RS) technology were used to create
    habitat models that related species presence and
    landscape characteristics. Habitat models were
    created for cougar, as well as wolf, deer and
    elk, for four different winter periods. Logistic
    regression and Akaike Information Criterion (AIC)
    were used for model construction and model
    selection respectively. Model accuracy was
    assessed using the Receiver Operating
    Characteristic (ROC) curve. Model predictions
    were used to describe cougar habitat suitability,
    as well as to infer the degree of competition and
    resource partitioning between the two top
    carnivores over the winter season. Cougar habitat
    selection was shown to be variable over the
    winter season, selecting for rugged regions of
    high elevations in early winter, while late
    winter selection indicates a shift in selection
    for areas lower on the hillslope, skirting the
    valley bottoms. Results indicate that competition
    and resource partitioning between cougars and
    wolves likely increases over the winter season,
    as predicted cougar distributions show increased
    overlap with the other species as winter

Julia LoVecchio
Site analysis for wind turbines
  • This study employs the spatial analysis and
    decision support functionality of geographic
    information systems (GIS) to select the most
    appropriate site for the installation of wind
    energy at the Sunshine Village Ski Resort in
    Banff National Park, Alberta. The study uses
    environmental features, such as those included as
    valued ecosystem components under the Canadian
    Environmental Assessment Act (CEAA), as
    constraints to the placement project, and
    identifies sites for turbine(s) in an area with
    winds exceeding average rates of 5 metres per
    second (m/s). Results indicate that 11 sites are
    available to the ski area for further analysis
    and investigation. None of these sites is located
    further than 500m from an electrical grid tie-in
    nor are they visible from the Trans-Canada
    Highway, Town of Banff, or the other ski areas in
    the National Park. Using one full-year of derived
    wind data from weather stations within the ski
    area and on Rundle Mountain, an economic analysis
    of the wind energy project was undertaken through
    the use of HOMER, a free-ware package created by
    the US National Renewable Energy Lab. Results
    indicate that the ideal turbine selection would
    be three Vestas V47s, 660kW systems producing
    roughly 6,420,305kWh of energy annually and
    reducing the ski resorts carbon emissions by
    over half. The wind mapping procedure used in the
    project is a simplified multivariate regression
    equation that reduces a complex physical
    phenomenon to three elements namely slope,
    aspect and elevation. Should the ski area choose
    to move forward with installing wind energy, a
    more sophisticated wind assessment will be

Nathalie Lowry
Mining effects on Yukon boreal forest
  • The Keno Hill mining district in central Yukon is
    located in the Boreal Cordillera Ecozone and has
    been subject to mining activity for more than
    eight decades. Using a combination of historical
    sediment metal concentrations, benthic
    invertebrate data and temporal Landsat satellite
    imagery collected from the Keno Hill district, an
    assessment of the effect of metal contamination
    on the boreal forest was made. The sediment
    metals and benthic invertebrates were evaluated
    on the basis of a control/impact experimental
    design to show that metals were having an effect
    on the stream ecosystem. The Landsat images were
    evaluated to detect changes in the forest over
    time using pseudo-supervised classification and
    post classification change detection analysis.
    The two were then used to see if a connection
    between metal contamination and forest change
    could be made. The sediment metal concentrations
    were higher at the affected sites and remained
    consistently high over the study period. Overall
    there was no difference detected in the benthic
    invertebrate communities over time though the
    site with the highest metal concentrations
    consistently had the lowest number of individuals
    and taxonomic richness. There was a change in
    vegetation between the two Landsat images where
    mixed spruce forest increased at the affected
    sample sites and decreased at the control sites.
    While it can be shown that there is evidence of
    metal contamination and a change in forest
    classification, a link between the forest and the
    metals cannot be made without access to forest
    classification information prior to mining
    activity. As this is not possible, future work
    using a similar analysis at different mine sites

Melanie Luinstra
Howler Monkey Habitat / Hurricane Regeneration
  • This study examines a population of howler
    monkeys (Alouatta pigra) in Monkey River, Belize
    whose habitat was affected by a hurricane in
    October 2001. Habitat models were created and
    selected for both the pre and post-hurricane time
    periods, using logistic regression and Akaikes
    Information Criteria applied at two different
    scales. Remote Sensing (RS) and Geographic
    Information System (GIS) techniques were employed
    to derive the independent variables and create
    the models. It was found that the monkeys were
    selecting for areas further away from rivers at
    the smaller study scale, but closer to these
    types of areas at the larger scale, they were
    selecting for different types of vegetation in
    the different time periods, and that their
    habitat was less homogeneous after the hurricane.
    The spatial extrapolation of these regression
    models was less successful. However, it is
    possible that areas to the east in the smaller
    study site were more suitable after the
    hurricane. More data collection is needed over
    the larger area, and more scale appropriate data
    may be used for the smaller area to attain
    generalizable spatial models.

David MacDonald
Classification of mixed prairie pasture types in
southern Alberta using Landsat TM and RADARSAT
  • Having a means of accurately classifying pasture
    types over large areas is important to ranchers
    and to conservation organizations interested in
    preserving native prairie. Radar imagery has been
    shown to be complementary to optical imagery with
    respect to providing information about rangeland.
    The purpose of this study was to use satellite
    imagery to assess the differences in radar and
    optical digital number (DN) values for different
    mixed prairie pasture types in southeastern
    Alberta and to develop a means of classifying
    them based on these differences. The images used
    included a Landsat TM image (acquired July 25th
    1998), and five RADARSAT standard beam mode
    images (acquired in September of 1996 and May of
    1998). Four pasture types were assessed,
    including native, Russian wildrye, crested
    wheatgrass, and a pasture that was originally
    cultivated and seeded to Russian wildrye but has
    reverted to native vegetation (referred to as
    native/Russian wildrye). The native and
    native/Russian wildrye pastures were found to
    have significantly lower average DN values for
    the Landsat image than the crested wheatgrass and
    Russian wildrye. Normalized difference vegetation
    index (NDVI) values were low for all classes
    indicating that there was little live healthy
    vegetation at the time of acquisition (i.e. the
    grasses had already senesced). Average radar
    backscatter was found to be significantly higher
    for the seeded pastures compared to native, the
    largest differences being between native vs.
    Russian wildrye and native/Russian wildrye. The
    fact that the native/Russian wildrye class
    (cultivated but with native vegetation) had
    average radar DNs most similar to those of
    Russian wildrye indicated that the roughness of
    the soil surface was dictating backscatter. Using
    the Landsat image alone for classification led to
    poor separability of the pasture classes.
    RADARSAT images alone, or together, also provided
    poor separability. Using a combination of the two
    types of imagery greatly increased the
    separability of the classes. The S7 image
    acquired in September was found to be the most
    useful for distinguishing between native and
    seeded pastures when added to the Landsat image
    but the September S1 image was found to provide
    the best increase in separability between crested
    wheatgrass and Russian wildrye. The S7 image
    acquired in May proved to be the least useful
    with respect to providing separability between
    classes. A combination of the Landsat bands (3,4,
    and 5), an NDVI transformation, Frost filtered
    RADARSAT images, Mean texture measure for
    RADARSAT images, Angular 2nd Moment and Standard
    Deviation texture measure images run on both the
    September RADARSAT images and band 4 of the
    Landsat image provided the best overall
    separability of pasture classes. Unsupervised
    classification showed the influence of geological
    features on the imagery but failed to
    discriminate pasture types. For the supervised
    classification, overall accuracy for the optimal
    combination of imagery and transformations was
    84.0, Kappa hat 76.7 for training sites, and
    73.5, Kappa hat 61.9 for randomly selected
    points. Although this level of accuracy does not
    allow for automated quantitative measurement of
    pasture areas by type, the classification
    obtained provides a means of visually identifying
    pasture species type for pure paddocks of
    Russian wildrye and native but less so for
    crested wheatgrass (accuracy for this class was
    low). In summary, the combination of C-band radar
    imagery and medium resolution optical imagery
    allows reasonable separation of Russian wildrye
    pasture from native pasture. This tested
    combination was not able to adequately separate
    native from crested wheatgrass it is suggested
    that optical imagery at a different phenological
    stage or radar imagery with a smaller wavelength
    (e.g. X-band), or with multiple polarization
    might be more profitably employed.

Shilong Mei
Making features movable in ArcGIS A
functionality of paleogeographic reconstruction
developed using ArcObjects and VBA
  • In current GIS software and GIS applications,
    features, once digitized into a vector layer,
    have fixed present day locations in a selected
    coordinate system. To deal with paleogeographic
    information, there is a need to restore the past
    locations of features of interest. A few
    Paleocontinent Reconstruction software packages
    have been developed for geologists independently
    from GIS. These software packages do not have the
    powerful functionalities in data management, data
    display and spatial analysis of the conventional
    GIS software. Current GIS software packages,
    however, do not have the function to restore past
    locations of geological features, but do have the
    potential to realize it. Paleogeographic
    Information System (PGIS), proposed by Mei and
    Henderson (2002), is an approach to incorporate
    the Paleocontinent Reconstruction function into
    GIS. This approach has been realized in ArcMap by
    developing a Paleogeographic Reconstruction
    functionality using ArcObjects and Visual Basic
    for Application (VBA) embedded in ArcMap.

Sunghuan Moon
DSS traffic issues
  • Calgary is one of the fastest growing cities in
    North America. The growth is expected to continue
    until the middle of 2030. As the city continues
    to grow, many new public facilities will be
    necessarily needed over the next 30 years.
    Locating new schools, and in particular high
    schools for grades 10 12, is a significant and
    political issue in City of Calgary. Site planning
    of schools can be examined in the context of
    spatial modeling methodologies focusing on
    maximizing students accessibility to their
    schools and minimizing the accessibility costs
    based on different algorithms. To solve these
    various accessibility patterns, several location
    models and spatial analysis methods have been
    used in geography, operational research and other
    disciplines. This paper presents an integrated
    spatial solution to the school site problem by
    combining location-allocation modeling and
    spatial interaction methodologies. Student
    population projections are analyzed in various
    aspects before implementing these models. The
    location-allocation modeling is used for locating
    new high schools depending on various situational
    parameters such as efficiency, equity, number of
    schools, and scale of economies. The spatial
    interaction model is used to estimate the new
    school location sensitivity, that is, variability
    of accessibility if the new locations change. GIS
    is used to solve various complex spatial
    problems. Powerful and flexible spatial modeling
    integration within GIS can contribute to
    extending the potential scope of GIS toward not
    only determining school locations, but also other
    spatial analysis and modeling applications.

Rupa Mukherjee
Spatial Analysis of Meso-Carnivores and Prey in
the Rocky Mountains
  • Wildlife conservation in the Canadian Rockies is
    complex in nature, and present a challenge to
    conservationists. One method of understanding the
    dynamics of an ecological system is to model what
    is occurring in nature. Habitat suitability
    modelling attempts to replicate the wildlife
    habitat processes that occur in nature. Two goals
    were present within this study (1) To determine
    how the community species affect habitat
    suitability for the Canada lynx (Lynx canadensis)
    and (2) To determine changes in the use of lynx
    habitat over a period of three years using a
    Geographic Information System (GIS), coupled with
    logistic regression and AIC (Akaikes Information
    Criteria) statistical techniques. Species are
    part of a sophisticated network of community
    interactions, which introduces complexity in
    modelling the system. Three key species within
    the lynx community were identified for analysis
    Snowshoe hare (Lepus americanus), coyote (Canis
    latrans) and American marten (Martes americana).
    Hare is considered the most important source of
    prey for lynx (Buskirk and Ruggiero, 1994).
    Coyote and marten are regarded as common
    competitors of lynx for prey (hare) (Lachowski,
    1997 ODonoghue et al., 1998). All three species
    along with landscape attributes were incorporated
    into the lynx model for the three-year study
    period (1997, 1998 and 1999). Image comparisons
    between the three habitat suitability maps showed
    that models varied annually in their predictions
    of total high suitability habitat. All three
    species (hare, coyote and marten) were integral
    to the top models created for lynx in all three
    years. Northness, greenness and terrain
    ruggedness were found to be the most influential
    landscape attributes among the top lynx models.

Anthonia Onyeahialam
Assessment Modeling of Sage Grouse Breeding
Habitats in North Natrona, Wyoming A RS, SS and
Predictive Modeling Approach Towards
  • This project presents results from an exploratory
    approach studying sage grouse breeding habitat,
    leks, in North Natrona County, Wyoming USA, using
    spatial statistical analysis, predictive
    modelling and Geographic Information Systems
    (GIS). It was found that sage grouse breeding
    locations present a pattern derived from a
    specific habitat selection process. Contributing
    habitat predictor variables in the models were
    generated from Landsat imageries using advanced
    image processing methods such as hierarchical
    classification, spectral pixel unmixing,
    knowledge based classification etc. Further, GIS
    was used extensively in data generation and
    analysis. Four final models that describe sage
    grouse breeding habitat selection were developed
    for multiple scales using logistic regression and
    multivariate adaptive regression splines (MARS).
    AIC, AICc and model averaging were used to aid
    the model selection process in identifying
    relevant predictors for lek locations and
    choosing the most appropriate model. Two of the
    models developed on the home range scale improved
    these predictions. Based on a large set of 83
    candidate models, important habitat predictor
    variables in the final models were elevation,
    distance to human development, slope, distance to
    roads, NDVI and distance to water. Sagebrush
    vegetation was not found among the most important
    predictor variables. Finally, a cumulative
Write a Comment
User Comments (0)